CMS finalized the CY 2027 MA and Part D Rate Announcement on April 6, 2026. The net payment update is +2.48%, roughly $13 billion in aggregate plan payments. That is up from the +0.09% / $700 million figure CMS published in the January Advance Notice. The swing was driven by CMS abandoning its proposed transition to the 2024-calibrated risk adjustment model. CMS will keep the 2018-diagnosis / 2019-expenditure calibration for CY 2027.

The part of the rule that matters more for physician groups is buried in the methodology section. CMS finalized the proposal to exclude unlinked chart-review diagnoses from risk score calculations. CMS estimates this exclusion alone will save $7 billion in CY 2027.

What "unlinked chart review" actually means

A chart review diagnosis is one captured by retrospectively reviewing patient documentation, often months after the encounter. An "unlinked" chart review diagnosis is one that has no underlying encounter, lab, imaging, or treatment activity tied to it in the claims data. The diagnosis appears in the risk adjustment submission but nowhere else in the patient's clinical activity for the year.

For about ten years, this has been a major capture lever for MA-aligned medical groups. Coders review charts, find documented conditions not previously submitted, add them retroactively. The plan submits them. Risk score goes up. Revenue follows.

Starting CY 2027, that capture line goes to zero. Any diagnosis that cannot be tied to an encounter, lab, imaging, or treatment activity is excluded from risk score calculations.

The two opposing signals for physician groups

Payment is up. Scrutiny is up. Plans will have more dollars to spend, which should ease the squeeze on shared-savings and capitation deals that got tightened in 2026. But any RAF lift your group books from a retrospective chart sweep without encounter linkage is dead money starting CY 2027.

That is the operational model many physician groups built since 2020. The annual cycle of AWV plus retrospective sweep, with the sweep contributing 30 to 50 percent of total HCC capture. The sweep half of that model is structurally dead in 2027.

What the math looks like at the group level

A practice with 8,000 MA lives and a typical chart-sweep contribution of 0.08 RAF points is now leaving roughly $1.4 million annually on the table starting January 2027. The fix is to migrate capture upstream, into the encounter itself, where every coded condition has a concurrent clinical event attached.

The practices that already operate this way (real-time NLP at the point of documentation, MEAT-criteria check before claim drop, clinician-attested coding) are net winners. The practices that still rely on quarterly chart-sweep cycles need a new capture model in place by Q4 2026.

What this means for your practice

Pull your last 12 months of HCC capture and sort by source. Separate encounter-linked captures from chart-review-only captures. If chart-review-only is more than 20% of your total HCC capture, you have a CY 2027 revenue cliff to plan around.

The migration to encounter-linked capture is technically straightforward, operationally hard, and politically charged inside the clinic. Start the design now and run the change management before the CY 2027 plan year opens.

✓ This week

Pull your last 12 months of HCC capture. Sort by source. Encounter-linked versus chart-review-only. Compute the chart-review-only percentage. Then model what your CY 2027 RAF score looks like if that revenue line goes to zero.

What to watch next

CMS will publish the CY 2027 Rate Announcement Technical Notes in late summer 2026. The notes will specify exactly how the unlinked-chart-review exclusion is operationalized in submission processing. Plans will then update their delegation arrangements with physician groups. Expect contract renegotiations in Q4 2026 as plans rebalance shared-savings math around the new capture rules.

ASP-RCM does this

HCC AI Engine v3.1 enforces encounter linkage at code suggestion time.

Every captured HCC carries a dated encounter timestamp. Every code suggestion runs a MEAT-criteria check with documented clinician override. The audit trail is RADV-defensible by design. Customers running v3.1 are seeing an average RAF lift of 0.14 in 90 days, all encounter-linked.

Book a 30-minute working call →